TOPPAN develops AI OCR to read difficult-to-decipher “medieval Greek,” leveraging technology for deciphering Japanese “kuzushiji”

TOPPAN Holdings and TOPPAN unveiled an AI-OCR engine that can decipher medieval Greek manuscripts.

The system adapts image-recognition technology originally developed for reading Japanese cursive historical documents.

It was trained using manuscript data from the Vatican Apostolic Library, with accuracy improvements aimed at 95% or higher.

A demo will be shown at an exhibition at the Printing Museum in Tokyo.

The technology could help automate the digitization, preservation, and research use of hard-to-read historical texts.